Sampling strategy by statistical analyses of LIBS data in the context of a portable gold analyser

摘要

Canada's Quebec province operates 26 metal mines, 19 of which are producing gold and generating over C$1.8 billion (US$1.3 billion) annually. In all those mines, numerous samples of rock are collected and analysed daily in laboratories in order to determine their gold content. However, laboratory results are generally available within 48 hours of sampling (depending on the number of samples received by the laboratory and the priority of the results), thus potentially slowing down mining activities. In an attempt to find a solution for delayed laboratory results, Laval University, in partnership with the National Research Council Canada (NRC), the National Institute for Scientific Research (INRS) and four mining companies (Agnico Eagle, Canadian Malartic, Hecla Quebec and Iamgold) have launched the Laser-Induced Breakdown Spectroscopy (LIBS) project for in situ real-time mining sample analysis. This work aims to minimise the amount of LIBS point sampling necessary to be representative of a full scan of the analysed area. Eight sampling patterns are investigated. A statistical analysis of the data was performed by varying the pattern steps. Results show that two of the patterns, 1 and 5 (single diagonal crossing the stratification of the samples with opposite slopes) minimise the amount of point sampling while remaining representative. The most successful results are obtained from steps which are greater than 6. In terms of the type of material, pattern 5 and steps 9 are better for core sampling, while pattern 1 and steps 8 are better for rock sampling. The geological explanation for these results is the heterogeneous distribution of gold in the sample which seems to be more homogeneous in the rocks than in the core sample due to the small rock size. In terms of the mineralisation, pattern 5 is better for mixed sulfides and quartz veins, and steps 8 and 9 are better for quartz veins.